July 28, 2017
Assessment of Landscape Changes in the North Atlantic
Landscape Conservation Cooperative: Decision-Support
Tools for Conservation (Phase 1)
This document is intended to serve as a general description of this project and, more
specifically, provide an overview of our approach to meeting the goals and objectives.
This working document is directed to our Scientific Steering Committee, but should be
useful to anyone interested in learning more about this project. This document is
something of an executive summary of the project in its current state and makes
reference to other documents (some complete, but mostly under development) that
provide the technical details of the approach.
1. Goals and objectives
Our primary mission as conservationists and public stewards of fish and wildlife
resources is to ensure the conservation of biological diversity. Thus, our primary overarching goal is to maintain well-distributed viable populations of all native species and
the ecosystem processes they perform and depend on. To achieve this goal, however, we
face many serious challenges associated with human population growth, such as habitat
loss and fragmentation, disruption of ecological processes, spread of invasive non-native
species, and human disturbance, all of which are being overlain and exacerbated by
global climate change. In the face of these serious challenges, our conservation objective
is to maximize the quantity, quality,
and connectivity of habitats and
ecological systems, subject to the real
world socio-economic constraints of
development. More specifically, our
conservation objective is to protect,
manage and restore as much habitat
as possible, minimize the forces of
habitat degradation, and design
landscapes to ensure habitat
connectivity within the limits imposed
by the socio-economic realities of
human population growth.
To achieve this conservation
objective, the USFWS developed the
Strategic Habitat Conservation
(SHC) approach, which incorporates
five key principles in an ongoing
process that changes and evolves in
an adaptive framework (Fig. 1):
1|Page
 Biological Planning (setting targets)
 Conservation Design (developing a plan to meet the goals)
 Conservation Delivery (implementing the plan)
 Monitoring and Adaptive Management (measuring success and improving results)
 Research (increasing our understanding)
To implement the SHC approach, the USFWS created a geographic network of
ecologically-based Landscape Conservation Cooperatives (LCCs)(Fig. 2). The North
Atlantic LCC (NALCC) was established in 2010 and encompasses the mid and north
Atlantic coast, including all or part of 12 states from Virginia to Main (Fig. 3). The
specific goals of the NALCC are as follows:
1.
Assess the current capability of habitats to support sustainable populations of
wildlife;
2. Predict the impacts of landscape-level changes (e.g., from urban growth,
conservation programs, climate change, etc.) on the future capability of these
habitats to support wildlife populations;
3. Target conservation programs to effectively and efficiently achieve objectives in
State Wildlife Action Plans and other conservation plans and evaluate progress
under these plans; and
4. Enhance coordination among partners during the planning, implementation and
evaluation of habitat conservation through conservation design.
2|Page
The project described in this
document is one of the sciencedevelopment projects of the
NALCC. While the focus of this
particular project is #1 and #2
above, the modeling framework
being developed will provide the
basis for #3 in the long term.
Thus, the proposed modeling
framework will allow us to assess
landscape change, assess
changes in ecological integrity
and habitat capability for
representative species, and allow
us to identify priorities for land
protection (i.e., what lands to
protect to get the biggest bang
for the buck) and conservation
priorities for existing
conservation lands (i.e., what
should be the management
priorities on each conservation
tract). The specific objectives are
as follows:
1.
2.
3.
4.
5.
Develop a landscape
change, assessment and
design (LCAD) model for
the NALCC that will allow
us to predict changes to the
landscape under a variety of
alternative future scenarios (e.g., climate change, urban growth), assess affects of
those changes to ecological integrity (coarse filter) and habitat capability for
representative species (fine filter), and eventually (in phase 2) allow us to design
conservation strategies (e.g., land protection, management and resotoration) to
meet conservation objectives.
Develop habitat capability models for a suite of representative species to be used
as a fine filter for evaluating the landscape change scenarios above (#1). Note, this
objective is being met in collaboration with the University of Vermont.
Develop ecological integrity models for a suite of ecological systems to be used as a
coarse filter for the evaluating landscape change scenarios above (#1).
Pilot the LCAD model by simulating landscape change and the corresponding
affects on ecological integrity and habitat capability for the representative species
in three representative watersheds distributed throughout the NALCC.
Assess the nature and magnitude of differences and similarities between areas
identified as important habitat for the representative species (fine filter) and areas
identified as having high ecological integrity (coarse filter) within the pilot
3|Page
watersheds; describe the implications for strategic habitat conservation planning
and make recommendations for effectively combining fine- and coarse-filtered
approaches to habitat conservation.
Note, these project objectives dovetail tightly with the first two steps of the SHC
approach: (1) biological planning and (2) conservation design. Specifically, the LCAD
model will provide a landscape assessment tool that can inform biological planning and
a landscape design tool that can inform conservation design.
2. Model Design
The LCAD model is being designed with the following important considerations in
mind:
1. Computational feasibility.--The model must be practical to run given available
computing resources. This will involve simplifying the model as necessary so that it
is practical to run. A "good" model that we can run in days is better than a "great"
model that we need a super computer and a year to run.
2. Extant data.--The model input data must be based on extant data at the regional
scale or data that can easily be compiled at the regional scale, and the model
complexity must be scaled appropriately to match the quality of the data. We do not
have the time or resources to be developing raw data, so our model will have to be
limited to what already exists for the most part.
3. Minimize subjective parameterization.--The model should require as few subjective
parameters as possible, use empirically-derived parameter estimates wherever
possible and resort to expert opinion only when necessary. This has implications in
the choice of methods. For example, rather than use expert state-based transition
models for vegetation development (succession), we will opt to use FIA data driven
FVS models of continuous vegetation change.
4. Model uncertainty.--The model must allow us to explicitly examine uncertainty in
predictions (based on the uncertainty in model parameters). Note, assessing model
uncertainty comes at the great cost of additional computations, so there is a real
tradeoff between computational feasibility and modeling uncertainty -- we will strive
to strike a balance.
5. Fisheries project compatibility.--The model should strive for compatibility with the
fisheries project, particularly with respect to the spatial and temporal scale of the
model and the particular ecological attributes tracked in the model.
6. Coarse- and fine-filter assessment capability.--The model must provide a
framework for both species modeling and ecological integrity modeling.
7. Start simple.--The model should be kept as simple as possible at first without
compromising the ability to add complexity later as time, resources and knowledge
allow. For example, while we would like to incorporate a mechanistic model of the
relationship between climate and vegetation development, we will likely have to
adopt a much simpler indirect approach that involves modeling the migration of
ecological systems and their coupled vegetation development models.
4|Page
Given the considerations above, the broad LCAD modeling framework for the combined
NALCC (i.e., inclusive of the fisheries project) is illustrated in figure 4. Briefly, in
addition the spatial and non-spatial database, the model is conceptually comprised of
three major components (described below), including:
1. Landscape change -- This is the core landscape change model, where the landscape
drivers (e.g., urban growth) and vegetation succession processes are implemented
under a user-specified scenario or set of scenarios and user-specified number of
stochastic runs of each scenario. This is where the ecological setting variables (i.e.,
spatial data layers representing biophysical and anthropogenic attributes of the
landscape) are modified over time to reflect the landscape drivers and succession.
2. Landscape assessment -- This is the coarse-fine filter assessment of landscape
ecological integrity (coarse filter) and habitat capability for representative species
(fine filter) at each timestep and summarized for the simulation run and scenario as
a whole. This assessment is used to evaluate the ecological performance of a scenario
by comparison to the baseline starting condition and to each other, and is the basis
for informing landscape design.
5|Page
3. Landscape design -- This component involves designing a land protection, land
management, and/or restoration scenario to maximize ecological performance
criteria such as the landscape ecological integrity indices (coarse filter) and habitat
capability indices for representative species (fine filter).
The LCAD model is relatively straightforward in organization and implementation; it
involves iteratively implementing landscape design, change, and assessment processes
over timesteps for one or more scenarios, repeated many times to realize the
stochasticity of the model processes. The raw results are a set of settings grids and an
assessment of ecological integrity and habitat capability for each timestep for each
stochastic run of each scenario. The summary results are a set of grids depicting
ecological integrity, habitat capability and conservation priorities for each scenario and
a set of tables summarizing the landscape ecological integrity and landscape capability
indices for each representative species across scenarios.
The following is an outline of the modeling process; components in upper case are
collections of one or more computer functions related to a common purpose and are
briefly described in the following sections. Note, the outline below is for conceptual
purposes only; the actual technical implementation may be somewhat different as
needed for maximum computational efficiency.
1. Baseline assessment (timestep 0)
Coarse filter:
->Run INTACTNESS -- function to measure degree of freedom from human
impairment
->Run RESILIENCY -- function to measure capacity to recover from disturbance and
stress
->Run ADAPT -- function to measure capacity to adapt to a changing environment
driven by climate change and development
->Run IRA -- function to combine intactness, resiliency and adapt metrics into a
single composite metric
->Run DIVERSITY -- function to measure the diversity of ecological systems
represented in the undeveloped portion of the landscape relative to the target
diversity
->Run BUFFER -- function to measure the capacity for areas to help other highvalued sites retain their ecological value over time (i.e., buffer then from
sources of human impairment)
->Run CONNECT -- function to measure the propensity to conduct ecological flows
across the landscape
Result = suite of ecological integrity grids and numerical summaries of
landscape ecological integrity
Fine filter:
->Run HABIT@ -- functions to measure habitat capability for representative species
Result = habitat capability grid for each species plus landscape
capability index for each species
6|Page
Development:
->Run SPRAWL --function to create new development (used here only to produce
probability of each type of transition for use in SECURE in first timestep)
Result = probability of development grids (one for each transition
type)
2. Simulation
Loop thru stochastic runs 1-m:
Loop thru scenarios 1-k:
Loop through timesteps 1-n:
Landscape design:
->Run SECURE -- function to protect new land under a user-specified land
protection scenarios
Result = new secured land grid
->Run MANAGE -- function to actively manage vegetation under a user-specified
land management scenario
Result = new vegetation settings grids
->Run RESTORE -- function to actively restore ecological integrity under a userspecified restoration scenario
Result = new ecological and anthropogenic settings grids
Landscape change:
->Run GROW -- function to produce vegetation succession
Result = new vegetation settings grids
->Run SPRAWL --function to create new development
Result = new anthropogenic settings grids
->Run other disturbance processes (to be developed in phase 2) in random order.
Result =new abiotic and/or vegetations settings grids
Landscape assessment:
Coarse filter:
->Run INTACTNESS -- as above
->Run RESILIENCE -- as above
->Run IRA -- as above
->Run DIVERSITY -- as above
->Run BUFFER -- as above
->Run CONNECT -- as above
Result = suite of integrity grids and numerical summaries of
landscape ecological integrity
Fine filter:
->Run HABIT@ models -- as above
7|Page
Result = habitat capability grid for each species plus landscape
capability index for each species
->Run CLIMENV -- function to map climate niche envelop for each
representative species
Result = binary grid depicting contemporary climate envelope for
each species
->Run HABCLIM -- function to intersect habitat capability grid and climate niche
envelop grid for each representative species
Result = habitat capability grid with non-zero values within the
contemporary climate envelope for each species
End loop thru timesteps
End loop thru scenarios
End loop thru stochastic runs
3. Simulation summary -- suite of functions for generating tabular and graphical
summaries of the simulation; by timestep, run, and scenario
The model is entirely grid-based to facilitate modeling contagious processes (e.g.,
disturbance) and spatial dynamism in the environment. The spatial resolution of the
model is 30 m to be consistent with many of the input data sources. The temporal
resolution is 10 years with a temporal extent of 80-100 years. A 10 year resolution is
deemed a sufficient compromise between realistically representing processes that
operate at finer temporal scales (e.g., annual variability in climate) and vegetation
dynamics (e.g., seral stage changes) that are much slower, and the need for
computational efficiency. Note, however, that climate data will be compiled and
downscaled at a seasonal resolution for use in the hydrology and coupled fisheries
model, but it will be aggregated to the decadal resolution for use in the LCAD model
described here. Lastly, the model is designed to be run on sub-landscape tiles such as
watersheds (HUC 6-8) to allow for parallel processing at the regional scale and to
integrate well with the fisheries project, but should be flexible enough to work with any
geographic extent (e.g., to accommodate application-specific conservation planning
units).
3. Model Components
3.1 Input Data
The input data includes user defined scenarios comprised of both nonspatial parameters
(that control the simulation and the component processes) and spatial data (maps)
representing ecological settings variables and other ancillary variables.
3.1.1 Non-spatial run parameters
This represents tabular (nonspatial) input data used to control the model run, including
basic control over the length of the model run (i.e., number of time steps), the number
8|Page
of replicate runs, and which drivers to include. Here is where the user specifies whether
to run a single scenario or a range of scenarios to reflect uncertainty in the drivers. For
example, multiple scenarios might represent a range of estimates of climate change, sea
level rise and urban growth rates.
3.1.2 Non-spatial component parameters
This represents the tabular (nonspatial) input data used to control the individual LCAD
model component processes (e.g., succession and drivers); in other words, values for the
parameters that control landscape change, assessment and design. This consists of a
series of tables associated with each model component. The number and structure of the
parameters will vary among model components. For example, the succession
component includes a suite of parameters describing the growth function for each
vegetation variable (e.g., biomass) indexed by ecological system. For other components,
indexing by ecological system is not useful (e.g., urban growth), and the tables are
structured accordingly.
3.1.3 Spatial data (grids)
This represents the spatial (GIS) data used in the LCAD model and consists of ecological
settings variables and ancillary GIS data. A detailed description of the spatial data is
provided in a separate document (Spatial). Briefly, the ecological settings variables
include a parsimonious suite of static as well as dynamic abiotic and biotic variables
representing the natural and anthropogenic environment at each location (cell) at each
time step (Table 1). Static variables are those that do not change over time (e.g.,
elevation, incident solar radiation). Dynamic variables are those that change over time
in response to succession and the drivers (e.g., canopy cover, temperature, traffic rate).
Most of the settings variables are continuous and thus represent landscape
heterogeneity as continuous (e.g., temperature, soil moisture), although some are
categorical and thus represent heterogeneity as discrete (e.g., ecological system,
developed). Importantly, the settings variables include a broad but parsimonious suite
of attributes that can be used to define the ecological system at any point in time; they
are considered primary determinants of the ecosystem composition, structure and
function, and determine the ecological similarity between two locations. As such, they
play a key role in the coarse-filter ecological integrity assessment, they can be used in
species' habitat models to represent important habitat components, and can be used in
any of the landscape change model processes. Thus, the settings provide a rich,
multivariate representation of important landscape attributes.
In addition to the settings variables, the spatial database includes a variety of ancillary
data layers that may be used in any of the landscape change modules (e.g., the
calculation of individual ecological integrity metrics, downscaling climate, predictors of
urban growth, etc.), and to control the output of the analysis (e.g., to determine the
spatial extent of an assessment)(Table 2). Note, all of these ancillary data layers are
treated as static and thus do not change over time.
3.2 Landscape Change
9|Page
The landscape change model essentially includes two separate, sequential processes that
iterate over time (i.e., at each time step in the model) over the length of the model run.
3.2.1 Succession
The first process invoked in each time step is succession. Succession is modeled as a
deterministic change in a variety of vegetation attributes, including biomass, quadratic
mean stem diameter, stem density, and potentially canopy cover and canopy height
(depending on need in wildlife habitat models) according to a set of growth functions
established for each forested ecological system or group of similar forested ecological
systems. A detailed description of the succession model (GROW) is provided in a
separate document. Briefly, we used Forest Inventory and Analysis (FIA) plot data to
compute each of the vegetation variables and stand age of each FIA plot. Pooling across
all FIA plots within a forested ecological system or group of similar forested ecological
systems, we treated each derived vegetation variable (e.g., biomass) as the dependent
variable and stand age as the independent variable, and fit a nonlinear function (e.g.,
Michaelis-Menten, Monomolecular) using ordinary least squares estimation. This
process fit a function to the average growth trajectory. Thus, for any given stand age, the
growth function predicts the corresponding average vegetation settings value. If an
ecological system did not have a sufficient sample size (i.e., >50 FIA plots) or a sufficient
stand age distribution (i.e., range >50% simulation length), it was aggregated with an
ecological similar system. Non-forested systems are assigned an average value for each
vegetation attribute (pooled across FIA plots) and treated as static (i.e., constant over
time) in phase 1 of this project.
3.2.3 Drivers
Following succession, the landscape is subject to one or more “drivers” of landscape
change. Each of these drivers is modeled separately, either as a deterministic or
stochastic process, and acts differently depending on the settings variables; however,
they all act to modify one or more of the settings variables (table 1). Uncertainty in
deterministic processes (e.g., climate change) is accounted for extrinsically by running
multiple varying scenarios; uncertainty in stochastic processes (e.g., urban growth) is
intrinsic to the process itself (via random variables) and is addressed by running
multiple replicate simulations.

Climate change –climate change is modeled as a deterministic process by simply
downscaling the climate predictions associated with monthly temperature and
annual precipitation from an ensemble of global coupled atmospheric-ocean general
circulation models (AOGCMs). The uncertainty in climate change predictions stems
from using a range of standard emissions scenarios set by the Intergovernmental
Panel on Climate Change (IPCC) Special Report on Emissions Scenarios (SRES)
representing high, medium and low levels of predicted change. A detailed
description of the climate model (GCMd) is provided in a separate document. Briefly,
We used publicly available AOGCM data that had been downscaled to 1/8 degree
(approximately 12km) using the Bias Corrected Spatial Disaggregation (BCSD)
downscaling approach. We averaged the results of 16 AOGCMs to create an ensemble
average projection for each of 3 SRES scenarios, subtracted a baseline to create
projected anomalies, and resampled these data at 800m cells. We then combined
10 | P a g e
these data with 800m resolution, 30-year normal temperature and precipitation
data (PRISM Climate Group, Oregon State University) using the “delta method”.
These data were further resampled to 30m to match the other layers used in the
LCAD model. Note, climate change acts principally to modify the ecological settings
variables associated with temperature and precipitation, and thus causes each cell to
migrate through ecological settings space over time. Thus, the principal effect of
climate change in phase 1 of this project is in the landscape assessment of ecological
integrity (via the adaptive capacity metrics, see below) and habitat capability for
representative species (via the climate niche envelope, see below). In phase 2 of this
project, climate change may also modify the succession and disturbance models
applied to each cell.

Urban growth – urban growth is modeled as a stochastic process by predicting the
probability of each type of development at the cell level, determining the total
amount of development based on human population projections, and allocating the
development among types, all within local windows scaled to account for spatial
heterogeneity in development rates and patterns. A detailed description of the urban
growth model (GROW) is provided in a separate document. Briefly, urban growth is
modeled statistically based on historical land use data to derive a locally-varying,
relative probability of each of six types of development transitions (e.g., undeveloped
to low-, medium-, and high-density development, low-density to medium- and highdensity development, and medium-density to high-density development). Predictors
include a variety of spatial variables, including resistant kernels based on sources of
jobs (e.g., urban areas), transportation infrastructure, and suitability of land (e.g.,
slope, wetlands, soils, secured land). The result of this step is a surface depicting the
relative probability of each type of development for each timestep. The actual
amount of development at each timestep is determined by human population
projections corresponding to the SRES scenarios and allocated among transition
types according to observed distributions at the local scale (to be determined). The
uncertainty in urban growth predictions stems from the intrinsic stochasticity of the
process itself and is realized by running multiple runs of the same scenario, in
addition to the variation among scenarios that vary for example in predicted human
population growth. Urban growth will act principally to modify the ecological
settings variables associated with human development such as impervious, traffic
rates and development.

Sea level rise – in phase 2 of this project, sea level rise will be modeled as a
deterministic process in the same manner as climate change, but is contingent upon
the development of broadly acceptable sea level models by third parties.

Timber harvest – in phase 2 of this project, timber harvest will be modeled as a
stochastic process. The details of this process are not clear yet and remain to be
developed. However, we will probably harvest timber randomly (as opposed to a
deterministic schedule) on lands deemed eligible for timber harvest and restrict
consideration to even-aged silvicultural treatments only (to keep things simple).
Unfortunately, harvest policies vary among ownerships (industrial, non-industrial
private, state, USFS, NPS, etc.), among state agencies, among states, can change
11 | P a g e
radically in short amounts of time in response to economic and political winds.
Timber harvest, in terms of types of treatments and intensity of harvest, is extremely
unpredictable. This suggests the need for many scenarios. Our approach will allow
for complex spatial and temporal variation in management, but we will keep it
simple to start. Timber harvest will act principally to modify the vegetation settings
variables (e.g., biomass, canopy cover).

Agriculture development/loss – in phase 2 of this project, agriculture development &
loss will be modeled as a stochastic process. The details of this process are not clear
yet and remain to be developed. Agricultural development may be important in some
portions of the region. Shifting agricultural land use, for example shifting from
cropland to pasture, could be included, but is highly unpredictable. Agricultural loss
is more likely throughout the region and will be modeled as a probability of
agricultural land reverting to early-successional natural land. Agriculture
development/loss depends on the economy, soil suitability, urbanization, land costs,
taxes, and distance to markets and other factors. Given the complex nature of this
process, modeling agriculture development/loss is probably a low priority among the
list of drivers.

Natural disturbances – in phase 2 of this project, natural disturbances will be
modeled as a suite of stochastic processes using a common algorithm that simulates
initiation, spread, termination, and effects. There are several natural disturbance
processes under consideration, including the following:
o Fire – probably too rare to matter in the northeast (return intervals at the cell
level much longer than simulation length of 100 yrs), but may be more
important in the southern portions of the region.
o Wind – downbursts and tornadoes may be frequent enough in some portions of
the region (e.g., Adirondaks) to model; hurricanes may also be frequent enough
in some portions of the region to model, perhaps separately from downbursts
and tornadoes.
o Insects/pathogens – native insects and pathogens are largely endemic and
generally do not cause stand replacement; non-native invasive insects and
pathogens may be worth considering on a case by case basis. Hemlock wooly
adelgid may be worth modeling; spruce budworm is another possibility, but
unsure whether enough stand replacement occurs to warrant inclusion. Model
parameterization for any insect/pathogen disturbance is going to be extremely
challenging.
o Floods – ecologically important to riverine and riparian ecosystems, but largely
doesn’t cause stand replacement in riparian systems (perhaps due to regulation
of rivers via dams) and geomorphic impacts to streams and riparian areas,
while important, is too difficult to model.
o Beavers – important driver in riverine and riparian ecosystems; may be
possible to model.
o Storm surge/overwash – important geomorphic disturbance in coastal
ecosystems (especially barrier beaches), but too difficult to model.
o Others?
12 | P a g e
3.3 Landscape Assessment
The landscape assessment model includes two separate components to quantify and
map the distribution of ecological integrity (coarse filter) and habitat capability for
representative species (fine filter) over time under alternative landscape change
scenarios. The models are coupled with the landscape change model through the use of
the common ecological settings database.
3.3.1 Ecological integrity assessment (coarse filter)
Our coarse filter landscape assessment is based on the concept of landscape ecological
integrity and is described in detail in a separate document (INTEGRITY). Briefly,
Landscape ecological integrity refers to the ability of an area to sustain ecological
functions over the long term; in particular, the ability to support biodiversity and the
ecosystem processes necessary to sustain biodiversity over the long term. For our
purposes, an integral landscape has a green infrastructure (i.e., undeveloped lands)
containing a diversity of ecosystems with high intactness, resiliency and adaptive
capacity that are well-buffered from human stressors and highly connected. Based on
this definition, there are six key attributes of landscape ecological integrity; i.e.,
measurable attributes that confer ecological integrity either to the landscape as a whole
or to the site (cell) and thus, by extension, to the landscape as a whole, that are as
follows:

Intactness...refers to the freedom from human impairment (anthropogenic
stressors); it is an intrinsic attribute of a site (cell) that contributes to the ecological
integrity of the site itself and thus, by extension, confers ecological integrity to the
landscape as a whole. Intactness is measured using a weighted linear combination of
a broad suite of stressor metrics.

Resiliency...refers to the capacity to recover from disturbance and stress; more
specifically, it refers to the amount of disturbance and stress a system can absorb
and still remain within the same state or domain of attraction (i.e., resistance to
permanent change in the composition, structure and function of the system) (Holling
1973, 1996). Resiliency is an intrinsic attribute of a site that contributes to the
ecological integrity of the site itself and thus, by extension, confers ecological
integrity to the landscape as a whole. Resiliency is measured using a weighted linear
combination of two metrics: connectedness and similarity.

Adaptive capacity...refers to the capacity to adapt to a changing environment (e.g.,
as driven by climate change);in the context of LCAD, adaptive capacity reflects the
potential for adaptation via movement to and from a site in order to track favorable
conditions as they change over time under non-equilibrium dynamics. Adaptive
capacity is an intrinsic attribute of a site that reflects the ecological integrity of the
site itself and thus, by extension, confers ecological integrity to the landscape as a
whole. Adaptive capacity is measured using a single metric: adaptive capacity.
13 | P a g e

Ecosystem diversity...refers to the diversity of ecological systems (or diversity of
ecological settings) represented in the undeveloped areas relative to the targeted
diversity, which may be user defined or a function of the proportional representation
of ecosystems in the full landscape. Note, diversity is a collective property of the
undeveloped lands within the landscape and is not a measurable cell attribute.
Diversity is measured using a single metric: diversity.

Buffering...refers to the capacity to help other high-valued sites retain their value
over time (i.e., buffer them from sources of human impairment), or the degree to
which high-valued sites are buffered by lands of low or no risk of development; it is a
contingent-value attribute of a site (i.e., derives its value from its location relative to
other high-valued sites) that confers integrity to the landscape as a whole. Buffering
is measured using two closely related metrics, which are used in different contexts:
buffer and protect.

Connectivity...refers to the propensity to conduct ecological flows across the
landscape; it is a contingent-value attribute of a site (i.e., derives its value from its
location relative to other high-valued sites) that confers integrity to the landscape as
a whole. Note, connectivity here refers to the ability to conduct flows locally and
regionally. Connectivity is measured using a single metric: connectivity.
The ecological integrity assessment, consisting of quantifying the six attributes above, is
done at each timestep of the model and summarized for the entire run and across
stochastic runs for each scenario. The ecological integrity assessment is useful as a
means of comparing scenarios with regards to achieving biodiversity conservation, and
it is also useful as a basis for landscape design (see below).
3.3.2 Habitat capability for representative species (fine filter)
Our fine filter landscape assessment is based on the concept of habitat capability for a
focal species and is described in detail in a separate document (HABIT@). Briefly,
habitat capability refers to the ability of the environment to provide the local resources
(e.g., food, cover, nest sites) needed for survival and reproduction in sufficient quantity,
quality and arrangement to meet the life history requirements of individuals and local
populations. Habitat capability for a focal species is assessed using HABIT@, a multiscale GIS-based system for modeling wildlife habitat. HABIT@ is a spatially-explicit
model: the habitat value at each cell is dependent not only on the resources available at
that cell, but on resources available in the neighborhood (is there enough forage to
support an individual’s homerange?), on the configuration of resources (are they
juxtaposed or contiguous?), and on the accessibility of resources due to impediments to
movement (are food and nesting resources across a road from each other?).
In collaboration with the University of Vermont, a HABIT@ model is being developed
for each representative species. Note, the details of the model vary among species
depending on the species' habitat requirements, but include an assessment of the
availability of one or more local resources (e.g., nesting, cover, food) based on the
ecological settings database, summarized at the home range level, and indexed for the
landscape as a whole. These are largely expert models in phase 1 of this project.
14 | P a g e
Similar to the ecological integrity assessment, the habitat assessment, consisting of a
habitat capability assessment for each representative species, is done at each timestep of
the model and summarized for the entire run and across stochastic runs for each
scenario. The habitat assessment is useful as a means of comparing scenarios with
regards to achieving biodiversity conservation, and it is also useful as a basis for
landscape design (see below).
3.4 Landscape Design
The landscape design model accommodates three different types of design applications,
corresponding to three separate models:
1. Land protection prioritization (SECURE)
2. Land management prioritization (MANAGE)
3. Restoration prioritization (RESTORE)
3.4.1 Secured land growth
The secured land growth model is intended to algorithmically implement different land
protections scenarios. The range of scenarios has yet to be defined. However, one
"smart" strategy involves first identifying and protecting high-valued core areas (based
on the landscape assessment), then protecting buffers around the cores, and finally
protecting areas that facilitate connectivity among cores. The details of this model are
not clear yet and remains to be developed. However, we will probably build out
protected land under varying scenarios representing different levels of aggressiveness.
For example, we could simulate purely opportunistic (i.e., random) land protection; we
could use the indices of ecological integrity and/or habitat capability to target land for
protection; we could use risk of development to target cheaper lands (under the
assumption that land values relate to risk of development); we could use a combination
of these measures to maximize opportunity; etc.. Secured land growth will act
principally to modify the secured lands setting variable and will interact strongly with
the urban growth model (SPRAWL).
3.4.2 Land management
The land management model is intended (in phase 2 of this project) to algorithmically
implement different land management scenarios aimed at active vegetation
management. The range of scenarios has yet to be defined, but may include treatments
such as timber harvest, mowing, and burning designed to create or maintain certain
vegetation conditions (e.g., early successional forest), possibly limited geographically to
specific ownerships (e.g., federal lands, protected lands, industrial lands). Land
management activities will act principally to modify the vegetation settings variables.
3.4.3 Restoration
The restoration model is intended (in phase 2 of this project) for implementing
ecological restoration activities (e.g., ecosystem conversion, wildlife passage structure,
culvert removal) and modifying the settings variables accordingly based on a userspecified restoration strategy. The range of scenarios has yet to be defined, but may
include activities such as culvert upgrades, dam removals, wildlife passage structures,
15 | P a g e
etc., and could be limited geographically to specific ownerships (e.g., federal lands,
protected lands, etc.).
16 | P a g e
Table 1. List of potential ecological settings variables. A detailed description of these
variables is provided in a separate document (Spatial Data). Note, S = static variables
(i.e., constant over time); D = dynamic variables (i.e., change over time). The variables
are arbitrarily grouped into broad attribute classes for organizational purposes.
Biophysical
attribute
Biophysical
variable
Temperature
Growing season
degree-days [D]
Growing season degree days
Min. winter
temperature [D]
Minimum January air temperature
Water temp [D]
Mean annual water temperature
Solar energy
Incident solar
radiation [S]
Solar radiation
Moisture &
hydrology
Topographic wetness
[D]
Soil moisture, measured by a topographic
wetness index
Flow volume [D]
Flow volume, measured as the flow
accumulation or watershed size
Flow gradient [S]
Gradient (percent slope) of a stream
Tidal regime [D]
In coastal areas, the frequency, period and
depth of tidal flooding
Soil pH [D]
Soil pH
Soil depth [D]
Soil depth
Soil texture [D]
Soil texture (based on average grain size)
Water salinity [D]
Salinity
Substrate mobility
[S]
Realized mobility of the physical substrate,
due to both substrate composition and
exposure to forces that transport material
Chemical &
physical
substrate
Description
17 | P a g e
Biophysical
attribute
Physical
disturbance
Vegetation
Anthropogenic
Biophysical
variable
Description
CaCO3 content [D]
Calcium content of the soil and water
Wind exposure [D]
Measure of exposure to sustained high
winds
Wave exposure [D]
Measure of direct exposure to ocean waves
Steep slopes [D]
Percent slope
Dominant life form
[S]
Dominant life form (unvegetated,
herbaceous, shrubland, woodland, forest)
Canopy closure [D]
Percent tree canopy cover
Canopy height [D]
Height of the dominant canopy
Biomass [D]
Total amount of above-ground vegetation
biomass
Mean diameter [D]
Quadratic mean tree diameter
Basal area snags [D]
Basal area of standing dead trees (snags)
Traffic rate [D]
Log of the average number of vehicles per
day on roads and railways
Imperviousness [D]
Percentage of the ground surface area that is
impervious to water infiltration
Developed [D]
Indicator of development (0=undeveloped,
1=developed)
Aquatic barriers [D]
Degree to which culverts and dams may
impede upstream and downstream
movement of aquatic organisms
Terrestrial barriers
[D]
Degree to which roads and railroads impede
movement of terrestrial organisms
18 | P a g e
Table 2. List of potential ancillary spatial data. A detailed description of these variables
is provided in a separate document (Spatial Data). Note, all of these ancillary data layers
are treated as static and thus do not change over time.
Variable
Description
Ecological
systems
Grid depicting TNC ecological systems
Land cover/use
Grid depicting land cover/land use classes
Elevation
Grid depicting elevation
Slope
Grid depicting raw percent slope
Aspect
Grid depicting aspect in degrees
Topographic
position
Grid depicting topographic position index
Hillshade
Grid depicting classic topographic hillshade
Topographic
roughness
Grid depicting topographic roughness index
Precipitation
Grid depicting downscaled precipitation (one for each SRES
scenario and GCM model)
Predicted matrix
system
Hexagons with predicted matrix systems from random forest
Small patch
systems
Grid depicting small patch systems classes
Wetland systems
Grid depicting wetland systems classes
Hydrography
Grid depicting stream flowlines
Flow grid
Grid depicting flow direction for each cell based on FD8 algorithm
Point source
pollution
Point coverage depicting point sources of pollution (such as
permitted toxic discharges into streams, municipal and industrial
sewage plants, and underground storage tanks, which may vary
among applications)
Traffic rate
Line coverage depicting road traffic rates (log transformed) for
each road segment and required input for the road traffic (traffic)
metric
19 | P a g e
Dams
Coverage/grid depicting the location and size class of each
recorded dam
Fine-grained
imperviousness
Grid indicating impervious surface (0=pervious, 1=impervious)
Wetland
polygons
Polygon coverage depicting wetland polygons
Coastal
structures
Line coverage depicting jetties, groins, sea walls and any other
beach-hardened coastal structure
Salt marsh
ditches
Line coverage depicting salt marsh
Tidal restriction
points
Point coverage depicting potential tidal restrictions (road and rail
crossings on tidal streams)
Boat traffic grid
Grid depicting boat traffic rates
Beach
pedestrians
Grid depicting beach pedestrian intensity
Parcels
Grid depicting ownership parcels
Census data
Grid depicting population predictions
Populated places
Polygons depicting the boundaries of urban centers of varying
population sizes
Secured land
Grid depicting protected lands
Include
Grid indicating whether to calculate the index of ecological
integrity or not (0=no, 1=yes)
20 | P a g e